This study first used measurements to establish the association between the rainy season precipitation in the Yangtze River valley (YRV) and north China (NC) and the 850-hPa meridional wind, and then evaluated the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) models’ simulations of both the associations and precipitation amount. It is shown that there exists a statistically significant positive correlation in the June–July precipitation and wind gradient over the YRV, and in the July–August precipitation and wind over NC. These associations are robust at daily, monthly, and interannual scales. Although many models are found to be capable of simulating the associations, the precipitation amount is still quite inadequate when compared with observations, thus raising the issue of the importance of lower-level wind simulations.
It remains a big challenge for climate models to realistically simulate regional precipitation (Meehl et al. 2005; Randall et al. 2007). One source of uncertainty is that regional climate processes, such as monsoon processes, cannot be represented explicitly in models, partly because of limitations in scientific understanding or in the availability of detailed observations of some physical processes (Christensen et al. 2007; Randall et al. 2007). Therefore, the study of relationships between climatic parameters related to regional climate processes—for example, the association of the meridional wind at low level with precipitation related to regional monsoon processes over eastern China—would not only improve our understanding of regional climate processes but also provide a potential approach for model evaluation.
The rainy season over eastern China is closely associated with the summer rainbelt. The summer rainbelt exhibits a stepwise movement northward from May through August following the summer monsoon advance from south to north. The rainy season emerges over southern China in May, the Yangtze River valley (YRV) in June–July, and north China (NC) in July–August (Tao and Chen 1987; Samel et al. 1999; Wang and LinHo 2002; Ding and Chan 2005; Ding and Wang 2008). Although the summer rainbelt over eastern China is controlled by abrupt jumps of the western Pacific subtropical high (WPSH) (Tao 1963; Ding 2004; Su and Xue 2010), the advance of low-level wind plays an important role in the march of the rainbelt in eastern China (Wang et al. 1999; Zhao et al. 2007). For example, the southeasterly wind that originated from the southwest flank of the WPSH occupying the YRV symbolizes the end of the rainy season in the YRV (mei-yu) and the onset of the rainy season in NC (Zhu 1934). It is evident that the seasonal variation of low-level wind is a fundamental feature of a typical monsoon domain, as indicated by previous studies (Ramage 1971; Murakami and Matsumoto 1994). So, what role does the meridional wind at the low level, as an important component of the East Asian summer monsoon system (Tao and Chen 1987), play in the advancement of the rainbelt and its precipitation variability?
Earlier studies used the meridional wind to study the intensity of the East Asian summer monsoon and its associated precipitation change over eastern China (Zhang et al. 1996; Wu and Ni 1997; Wang et al. 2001a,b; Wang 2002; Liang et al. 2008). Because of the spatial inhomogeneity of the meridional wind and the differences in the timing and duration of rainy seasons among the regions across eastern China, it is not suitable to use a summer monsoon index defined by the meridional wind in a specific period (e.g., June–August) over a certain region to understand the relationship between summer monsoon intensity and rainfall in different regions of eastern China during their individual rainy seasons or eastern China’s rainy season (May–August). For example, a summer monsoon index defined using 850-hPa meridional wind just explains the correlation between monsoon intensity and mei-yu rainfall change over the YRV in June well (Wang et al. 2001b). Therefore, the study of the local meridional wind and rainfall in different regions over eastern China during their rainy seasons would help us to understand their relationship. The temporal evolution of daily precipitation over the YRV and NC and the overlying meridional wind from 1 June to 31 August suggests that they are highly correlated with each other (Fig. 1). In the present study, we investigate the association of precipitation in the YRV and NC with 850-hPa meridional wind during their rainy seasons at different time scales (daily, monthly, and interannual) in the current climate background, and also examine these associations based on Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) coupled climate models.
2. Data and methods
Observational daily precipitation data from 1958 to 2002 are extracted from a 752-station database developed by the China Meteorological Administration (http://cdc.cma.gov.cn/shuju). Among them, precipitation records from 26 stations in NC (34°–42°N, 108°–121°E) and 33 stations in the YRV (26°–34°N, 109°–122°E) were used in this study (Fig. 1). The daily and monthly 850-hPa meridional winds at the horizontal resolution of 2.5° latitude and longitude are taken from the 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis (ERA-40) dataset (Uppala et al. 2005). Modeled data are also used to assess the relationship between precipitation in the YRV and NC and the associated meridional winds. The monthly outputs of the 20 coupled general circulation models (CGCMs) used in this study are from twentieth-century climate simulations for the IPCC AR4 (see appendix for the list of the 20 climate models). These datasets were obtained from the Program for Climate Model Diagnosis and Intercomparison (PCMDI) Web site (http://www-pcmdi.llnl.gov/). Detailed information about the models and their outputs is available on this Web site. For consistency in the comparisons of simulations with observations, 850-hPa meridional winds in the models were interpolated onto the same resolution as the ERA-40 2.5° × 2.5° grid. In this study, June–July and July–August mean anomalies are defined by the deviation of the 1971–2000 climatological mean.
Wang et al. (2001a) defined a monsoon index as the difference of the area-averaged 850-hPa meridional wind between two regions (20°–30°N, 110°–140°E; 30°–40°N, 110°–140°E) to reflect the prevailing position (south or north) of the East Asian summer monsoon. The boundary at 30°N is chosen because the mei-yu rainbelt is usually quasi-zonally oriented near 30°N over eastern China (Ding 2004). This study focuses on the prevailing position of summer rainbelts over eastern China rather than East Asia, and it emphasizes monsoon front activities and the relationship between “local” meridional wind and rainfall over the YRV and NC. Thus, the ranges of the two regions used by Wang et al. (2001a) are modified as 20°–30°N, 110°–120°E and 30°–40°N, 110°–120°E. The variables VS, VN, and VS − VN represent the averaged 850-hPa meridional winds over the southern and northern regions and their wind gradient. The anomalies of VS and VN are defined as having regional summer monsoon strength. A large positive value means a strong summer monsoon. The meridional wind gradient VS − VN reflects not only the summer monsoon strength but also the position of the rainbelt associated with the interaction of the air mass between south and north. A wind gradient with a large value means the strong summer monsoon stays mainly in the YRV and vice versa.
3. Results and discussions
As shown in Fig. 1, the precipitation in the YRV (NC) increases gradually from early June to a peak exceeding 7 mm day−1 (4 mm day−1) between mid-June and early July (between early July and mid-August), and then abruptly decreases to 5 mm day−1 (3 mm day−1) after early July (mid-August). These peaks of daily precipitation are consistent with the rainy seasons in the YRV and NC revealed by previous studies (Ding 2004; Ding and Chan 2005). For the meridional wind, the southerly wind prevails over the YRV and NC with a gradual decrease from south to north, except the weak northerly wind occurs in mid-August over NC because of more cold air activities from mid- and high latitudes. The temporal patterns of VS and VN are generally parallel to those of precipitation in the YRV and NC, although this parallelism over NC in July is not as good as in August. Correlation coefficients between the precipitation in the YRV and the VS − VN and VS are 0.75 and 0.48, respectively. Both are significant at the 99% confidence level. They also indicate that VS − VN has a greater impact on the precipitation change in the YRV than the VS. For NC, the precipitation is remarkably correlated with the VN, as indicated by the coefficient of 0.5, significant at the 99% confidence level. In summary, it can be concluded that the VS − VN and VN are closely associated with the precipitation in the YRV and NC from 1 June to 31 July and from 1 July to 31 August in respect to the climatological mean.
The accumulated rainfall of ~380 mm (~275 mm) over the YRV (NC) from 1 June to 31 July (1 July–31 August) accounts for 29% (50%) of the total annual rainfall amount. Although discrepancies in the rainy season onset and withdrawal dates exist among many previous studies (Tao and Chen 1987; Samel et al. 1999; Wang and LinHo 2002; Ding and Wang 2008) because of the use of different definition criteria of the onset and withdrawal and different datasets, there is consensus that the rainy seasons in the YRV and NC are within June–July and July–August, respectively; June–July for the YRV and July–August for NC are thus used as their extended rainy seasons in the following discussions.
As mentioned above, the climatological mean of the precipitation over the YRV during its rainy season is highly correlated with the VS − VN, whereas it is correlated with the VN over NC. To determine whether these associations are also robust during the rainy seasons of the two regions on daily and monthly scales, comparisons of them using data of all individual days and months during the rainy seasons from 1958 to 2002 are conducted (Figs. 2a–d). In the YRV, correlation coefficients between its precipitation and the VS − VN are 0.52 and 0.76 on daily and monthly scales, respectively. Both coefficients are statistically significant at the 99% confidence level. In NC, correlation coefficients between its precipitation and the VN are also significant at the 99% confidence level, although the variance explained by the daily correlation coefficient is not large because of the existence of a large divergence, especially for a smaller amount of daily precipitation. These comparisons indicate that the VS − VN and VN can reflect the variation of precipitation in the YRV and NC during their rainy seasons at daily and monthly scales, respectively. Higher precipitation in the YRV and NC are associated with larger VS − VN and VN, respectively, and vice versa. Further comparisons between the precipitation in the YRV and NC and their associated meridional winds during their rainy seasons on the interannual scale are shown in Figs. 2e,f, respectively. All the time series in Figs. 2e,f exhibit clear interannual and interdecadal variability. There is an increasing trend in the YRV rainfall and VS − VN. The precipitation change in the YRV is positively correlated to the VS − VN with a correlation coefficient of 0.72, statistically significant at the 99% confidence level (Fig. 2e). For NC, the VN in July–August has a close relation to the concurrent NC rainfall, with a correlation coefficient of 0.65 exceeding the 99% confidence level (Fig. 2f).
It should be pointed out that the result is not sensitive to the choice of reanalysis data. Our findings from ERA-40 are confirmed by the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (Kalnay et al. 1996). We conducted the same analysis on the NCEP–NCAR reanalysis data as ERA-40. The results show similar significant correlations. Correlation coefficients on daily, monthly, and interannual scales between the VS − VN and precipitation in the YRV are 0.46, 0.46, and 0.63, respectively, while the correlation coefficients between the VN and precipitation in NC are 0.21, 0.56, and 0.56, respectively. All these correlations are statistically significant at the 99% confidence level. Therefore, our analyses indicate that the associations of the precipitation in the YRV with the VS − VN and precipitation in NC with meridional wind VN during their rainy seasons are robust and reliable on daily, monthly, and interannual scales.
The mechanisms for these robust associations seem related to the summer monsoon and to the position of the rainbelt associated with the interaction of air mass between south and north. Since the meridional wind gradient (VS − VN) is defined as the difference of meridional wind between south and north, its anomaly is determined by the configuration of VS and VN anomalies. A positive VS − VN anomaly could be configured theoretically by the following three types of anomalies of VS (ΔVS) and VN (ΔVN): (P1) ΔVS > 0 and ΔVN < 0; (P2) ΔVS > 0, ΔVN > 0, and ΔVS > ΔVN; and (P3) ΔVS < 0, ΔVN < 0, and |ΔVS| < |ΔVN|. A negative anomaly of VS − VN can be configured theoretically by the anomalies of VS and VN: (N1) ΔVS > 0 and ΔVN > ΔVS; (N2) ΔVS < 0 and ΔVN > 0; and (N3) ΔVS < 0, ΔVN < 0, and |ΔVS| > |ΔVN|. To figure out the configuration of the VS and VN anomalies for the precipitation anomalies in the YRV during its rainy season, the 10 wettest and 10 driest cases are chosen. Among the 10 wettest (driest) cases, the VS − VN anomalies in 8 (9) cases are positive (negative), consistent with our analyses mentioned above. Among the 8 positive cases, there are 2 cases (1993 and 1998) in P1, 5 cases (1969, 1983, 1991, 1995, and 1996) in P2, and 1 case (1997) in P3. Among the 9 negative cases, there are 2 cases (1960 and 1961) in N1, 6 cases (1958, 1963, 1978, 1981, 1985, and 1988) in N2, and 1 case (1972) in N3. Note that the configurations of the VS and VN anomalies relative to the 1981–2000 climatological mean instead of 1971–2000 are the same. Therefore, a wet YRV is most likely caused by the P2 configuration (ΔVS > 0, ΔVN > 0, ΔVS > ΔVN), whereas a dry YRV is mainly caused by the N2 configuration (ΔVS < 0, ΔVN > 0). This feature implies that the YRV experiences a stronger summer monsoon in a wet year than a dry year and the mei-yu belt dominates the YRV during wet years. A stronger summer monsoon brings more warm/wet airflow into the YRV and thus produces more precipitation over there as the mei-yu front also occupies the YRV. This result needs verification by further analysis of the meridional vapor transport. For NC, no significant correlation on any scale exists between NC rainfall and the meridional wind gradient during its rainy season. The latter is defined by the difference of the area-averaged 850-hPa meridional wind between NC (30°–40°N, 110°–120°E) and the area north of NC (40°–50°N, 110°–120°E). This fact indicates that the meridional wind gradient in NC cannot reflect the interaction of air mass between south and north well as that in the YRV because of the topographic block effect of Yanshan Mountain and Taihang Mountain. Apparently, it is the summer monsoon strength that can be used to explain why the precipitation over NC is positively correlated to the local meridional wind VN during its rainy season.
Generally, climate models cannot simulate the East Asian precipitation well (Kang et al. 2002; Wang et al. 2004; Lin et al. 2008). Most of them fail to reproduce the East Asian summer rainbelts. The robust associations of rainfall over the YRV and NC and the overlying meridional winds revealed by our analyses reflect not only the prevailing position of rainbelts but also the summer monsoon strength, so they could provide a potential validation technique for the diagnosis of CGCMs. The three scatter diagrams in Fig. 3 show the correlation coefficients of modeled associations, observed and modeled regional low-level meridional wind, and observed and modeled regional precipitation in the YRV and NC during their rainy seasons based on 1958–99 data from 20 IPCC AR4 climate models. Among them, 17, 14, and 12 models can capture the observed association of precipitation over the YRV with VS − VN, the association of precipitation over NC with VN, and both associations (Fig. 3a), respectively. However, only models that also have good capability in simulating the observed meridional wind feature can replicate the regional precipitation in the YRV and NC better, for example, FGOALS-g1.0 (solid circle in Fig. 3). The simulated associations of this model are significant at the 99% confidence level with the correlation coefficients as high as 0.71 and 0.66 for the YRV and NC, respectively. Two coefficients have the third highest values among the 20 models, but this model produces the highest correlation coefficients between the observed and modeled regional precipitation in the YRV and NC during their rainy seasons among the 20 models because of its capability in simulating the observed meridional wind feature well (Figs. 3b,c). As shown in Fig. 3b, the correlation coefficient between the observed and modeled VS − VN for this model is significant at the 95% confidence level and the highest among the 20 models, while the correlation coefficient between the observed and modeled regional precipitation in the YRV during its rainy season is also the highest and the significant level is close to the 95% confidence level. This scatter diagram also indicates that the higher the correlation coefficients between the observed and modeled VS − VN are, the higher the correlation coefficients are between the observed and modeled regional precipitation over the YRV during its rainy season. This positive correlation is statistically significant at the 95% confidence level, suggesting that the better performance of the FGOALS-g1.0 model is not an occasional phenomenon. For NC, the scatter diagram (Fig. 3c) also shows such a relationship, although it is not significant at the 95% confidence level. A comparison of the modeled and observed climatological mean precipitation amounts over the YRV and NC during their rainy seasons of 1958–99 shows that the FGOALS-g1.0 has the smallest difference. Its modeled amount is 8.4 mm less than the observed one over the YRV, whereas it is 12.2 mm more than the observed one over NC. Among the 20 models, only FGOALS-g1.0’s and HadGEM1’s simulated amounts of both regions are within the 95% confidence intervals of their observed amounts. Note that these two models reproduce the observed associations’ precipitation with overlying meridional winds well and have better representation of the meridional wind feature than other models. Our results suggest that the model would most likely capture the observed precipitation feature if a model can reproduce both the observed associations of precipitation with overlying meridional winds and the observed meridional wind features during rainy seasons.
Our results indicate that there exist a significant positive correlation between the precipitation over the YRV and the 850-hPa meridional wind gradient between south and north (VS − VN) during its rainy season (June–July), and a significant positive correlation between the precipitation over NC and its “local” meridional wind VN during its rainy season (July–August) in 1958–2002. These associations are robust and reliable on daily, monthly, and interannual scales within their rainy seasons.
A comparison between the observations and simulations suggests that a good model at simulating the regional precipitation in the YRV and NC during their rainy seasons should meet the following two conditions: (i) a good relationship between the precipitation and low-level meridional wind and (ii) a good capability in simulating the observed meridional wind feature. It thus raises the issue of the importance of low-level wind simulations to the improvement of regional precipitation simulations.
This research was supported by one grant (to SUNYA) from the Office of Science (BER), U.S. Department of Energy; one grant (2010CB950100) from China Global Change Research Program; two grants (40905045 and 41071029) to NUIST from the Natural Science Foundation of China; a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD); and one grant (IAP09312) from the Knowledge Innovation Program of the Chinese Academy of Sciences. GZ was a visiting scientist (at SUNYA) under the joint agreement on “Climate Science” between the U.S. Department of Energy and the China Ministry of Sciences and Technology. We thank the two anonymous reviewers for their constructive suggestions that have improved our manuscript.
BCCR-BCM2.0 Bjerknes Centre for Climate Research Bergen Climate Model, version 2
CCSM3 Community Climate System Model, version 3
CGCM3.1-T47 Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model, version 3.1
CGCM3.1-T63 Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model, version 3.1
CNRM-CM3 Centre National de Recherches Météorologiques Coupled Global Climate Model, version 3
CSIRO Mk3.0 Commonwealth Scientific and Industrial Research Organisation Mark, version 3.0
ECHAM5/MPI-OM European Centre-Hamburg Model version 5/Max Planck Institute Ocean Model
FGOALS-g1.0 Flexible Global Ocean–Atmosphere–Land System Model gridpoint version 1.0
GFDL CM2.0 Geophysical Fluid Dynamics Laboratory Climate Model, version 2.0
GFDL CM2.1 Geophysical Fluid Dynamics Laboratory Climate Model, version 2.1
GISS-EH Goddard Institute for Space Studies Model E-H
GISS-ER GISS Model E-R
HadCM3 Third climate configuration of the Met Office Unified Model
HadGEM1 Hadley Centre Global Environmental Model, version 1
INM-CM3.0 Institute of Numerical Mathematics Coupled Model, version 3.0
IPSL CM4 L’Institut Pierre-Simon Laplace Coupled Model, version 4
MIROC3.2(hires) Model for Interdisciplinary Research on Climate 3.2, high-resolution version
MIROC3.2(medres) MIROC3.2, medium-resolution version
MRI CGCM2.3.2 Meteorological Research Institute Coupled General Circulation Model, version 2.3.2
PCM Parallel Climate Model